Gaussian Mixture Model (GMM) Cohort Analysis of Eastern Oysters in Chesapeake Bay
VOSARA Reefs
The map below shows the reefs in Virginia Chesapeake Bay monitored by VOSARA. Reefs in green are included in the cohort analysis, meaning for at least 8 consecutive years there were at least 300 oyster sampled. 64 reefs were included in this study, located in every main tributary except Lynnhaven Bay.
Chesapeake Bay Mainstem Reefs
The following map shows reefs in the Chesapeake Bay Mainstem. Reef 312 (Blackberry Hang) and Reef 357 (Deep Rock #1, Deep Rock #2, and Deep Rock #3) were included in this cohort analysis.
Gaussian Mixture Model Output
The GIFs below show the results of the Gaussian mixture model (GMM) used as a cohort analysis using the R function mclust::Mclust (Scrucca et al. 2023). Each age group was estimated using a GMM. Gaussian distributions were fit to the shell sizes (mm) of live oysters, and log-normal distributions were fit to the shell sizes (mm) of spat oysters . Those age groups (individual distributions) were linked and tracked using each mean estimate (\(\hat{\mu}\)). Each color represents a different cohort, shown by the legend in the top right corner. The asterisk over some curves represents insufficient data (n < 250 for lives, n < 50 for spat).
Both GIFs show a successful attempt to identify, link, and track age cohorts of eastern oysters in Chesapeake Bay. Both reefs show a noticeable signal that age-truncation issues (i.e., short lifespans) are improving: age groups are living longer in the past decade. Further, in recent years (2020 onward), there are more age groups present each year. Most notably, in 2022 and 2023, there are 4 age groups alive at one time in Reef 357.
Reef 312: Blackberry Hang
Reef 357: Deep Rock #1, Deep Rock #2, and Deep Rock #3
The patterns shown above are observed for the majority of reefs in the study, and also at the bay-level as a whole.
References
Scrucca, Luca, Chris Fraley, T. Brendan Murphy, and Raftery Adrian E. 2023. “Model-Based Clustering, Classification, and Density Estimation Using Mclust in R,” March. https://doi.org/10.1201/9781003277965.